A new approach to quantifying bodies of text is reviewed. The core of the software involved (CATPAC) is a self-organizing neural network. It begins with a set of artificial neurons; one for each word in the text it is reading. The analysis is initiated by passing a “scanning window" of n consecutive words through the text. The structure of the artificial brain thus established can be represented by a square matrix of numbers. Each row and column represents a neuron (word), while each number (an updatable weight) represents the strength of connections of the neurons corresponding to the row and column of the number (cell entree). The resulting matrix resembles a typ ical covariance or correlation matrix and can be used as input matrix for multivariate statistical analysis. A small illustrative study involves a focus group dealing with a holiday cruise. Results are scrutinized using cluster analysis and MDS. It is concluded that the software can be of value for the tourist researcher, provided that appropriate time is dedicated to the analysis.
|Status||Udgivet - 1999|